The last time I dared to dabble in dictation software was about 15 years ago, and even allowing for how that software only ran on a Windows box, the whole experience was extraordinarily cumbersome. But after using Dictate for only five minutes, the improvements in the technology since then are starkly apparent. It’s not perfect, but the dream of quickly turning spoken words into editable text has certainly progressed.
Dictate first walks you through its initial training, which we finished surprisingly quickly. After that, the software is immediately usable. In fact, all the text that you’ve read so far in this article--and much of the text that follows--was dictated right after completing the basic training process.
Dictate includes a straight dictation mode, a spelling mode, and a command mode, which is intended primarily to control the standard Mac UI via voice--important for anyone who can’t use a keyboard or mouse.
MacSpeech Dictate keeps you well informed about what it thinks you just said.
Dictate shines if you’re speaking straightforward text. However, English is loaded with aural ambiguity that can confuse humans, not to mention software. As testing progressed, Dictate’s confusion between plurals and possessives became irritating, though its learning abilities will likely reduce this annoyance over time. While in many cases Dictate’s onscreen recognition palette provided reasonable choices that could be selected verbally, other times we had to decide between a somewhat unwieldy verbal-editing process of selecting and replacing errors… or simply giving up and using the keyboard.
And trying to write and edit with a combination of speech and keyboard isn’t the way to go because much of Dictate’s understanding of a sentence comes from context. It doesn’t know when to capitalize sentences when you use it to simply insert patches of text, for example. But if you dictate text as a flow and specify punctuation and paragraph endings, Dictate does a solid job.
Dictate’s admirable ability to recognize large swaths of text at a time is also a double-edged sword because backtracking to verbally edit individual words or punctuation takes effort. While the initial temptation is to just carry on and fix the mistakes later, sometimes the misrecognized words form strange enough non sequiturs that they’ll be incomprehensible later. But editing on the spot also derails your train of thought, which negates the benefits of continuous speech recognition. The ideal of being able to pace around the room, dictating pages upon pages of text hasn’t quite arrived… yet.
Accurate, continuous recognition. Works with every app we tried.
Easily confuses plurals and possessives. Verbal editing unwieldy. Not available without bundled microphone.
